Grammar elements in Nylon act as natural language representations of machine learning tasks. For example, the keyword "one-hot", allows you to One-Hot Encode a column. We hope that this simpler way of interacting with different ML components, make it easier for you to customize your workflow.
To be clear, this is the grammar that will be specified in the specifications file provided to your training instance.
This guide for the grammar is split up into several components -- each representing a different part of the machine learning pipeline. These processes work independently, and so, a full understanding of Nylon's grammar is not necessary. Anything you don't want to spend time understanding, you can just let us handle.
You can find the different vocabulary elements you'll be able to use as a part of Nylon on the page below. Each section concerns a different part of the ML pipeline.
Each section below works under a different tag in the specifications file. The first, is preprocessing, which can be modified under the preprocessor tag in the grammar.
Next is modeling, which works under the modeling tag in the grammar. The different options for modeling are provided below.
Analysis is the final section, which works under the analysis tag in the grammar. Analysis blocks can also be used to modified how the results of your model are evaluated.
More granular customization will be coming soon.